Family caregivers for cancer patients experience high levels of stress and burden and diminished quality of life (QOL). Interventions to improve coping skills of caregivers have been shown to be effective with other populations, but their impact has not been assessed in the difficult context of hospice care. The purpose of this study was to determine whether hospice plus a coping skill training intervention improved family caregivers' QOL, burden, coping, and mastery, compared with hospice plus emotional support, and usual hospice care.
A three group randomized controlled trial was conducted including baseline, 16 day, and 30 day assessments conducted from March 1999 to May 2003. The sample consisted of 354 family caregivers of community dwelling hospice patients with advanced cancer. Patient/caregiver dyads were randomly divided into three groups, including a control group (n = 109) who received standard hospice care, a group (n = 109) who received standard hospice care plus three supportive visits, and a group (n = 111) who received standard care plus three visits to teach a coping skills intervention. Primary outcomes included caregiver QOL, caregiver burden due to patient symptoms, caregiver burden due to tasks, and caregiver mastery.
At the 30-day follow-up, the coping skills intervention led to significantly greater improvement in caregiver QOL (estimate = −0.16, standard error [SE] = 0.07, P = 0.03), burden of patient symptoms (estimate = 0.28, SE = 0.07, P < 0.001), and caregiving task burden (estimate = −0.01, SE = 0.01, P = 0.038) than did the other two conditions. None of the groups showed significant change in overall caregiving mastery, caregiver mastery specific to caregiving tasks, problem-focused or emotion-focused coping.
Up to 52 million Americans provide assistance to a family member with an illness or disability,1 and many caregivers provide extensive assistance to impaired relatives with cancer. Highly stressed family caregivers have been found to be at increased risk for depression and health problems2–3 and increased mortality rates.4
Research to date has suggested that, while continuing to face heavy caregiving stress, caregivers can benefit greatly from structured psychosocial interventions that teach coping and problem-solving skills. A recent metaanalysis5 provides evidence that such caregiver interventions can lead to clinically significant improvements in the well being of the caregiver. To date, most of this research has been conducted with family caregivers of persons with Alzheimer disease and other dementias, although there is limited literature demonstrating the benefit of caregiver interventions in cancer.6–8
One group of family caregivers that experiences very high levels of stress, but has been studied little, is family members providing care for terminally ill cancer patients. Research has demonstrated that such caregivers experience high levels of burden due to intense physical, psychological, and economic strain,9–10 facing not only stresses of caregiving but also the impending death of a loved one. Even in the context of hospice, where extensive supportive services are available, family caregivers of terminally ill cancer patients report providing an average of 115 hours of care each week and report depressive symptoms and health problems far in excess of age-matched controls.11 Caregiving for terminally ill cancer patients also takes a toll on quality of life (QOL),12 and can erode the caregiver's sense of mastery.13
Decreasing the burden on family caregivers and improving their QOL have been widely recognized as important goals of hospice and palliative care,14–15 although there is currently an insufficient evidence base to guide care for families providing care at the end of life.16 Family caregivers who receive care from hospice report high satisfaction with hospice care,17 but even caregivers who benefit from hospice care have unmet needs due to the nature of caregiving demands at the end of life.
Our literature review has identified no randomized intervention studies examining the benefit of structured psychosocial interventions for caregivers of hospice patients with cancer. On the basis of positive results from interventions for other groups of caregivers, we adapted an existing problem-solving intervention that has been used with cancer patients18 and family caregivers of cancer patients19 for use in the hospice context. Because median length of stay in hospice is only 20.9 days20 and caregivers are under tremendous strain during this period, we developed and assessed the impact of a problem solving intervention on hospice caregivers' QOL, burden, mastery, and coping. We predicted that a brief problem solving intervention, provided in addition to hospice cancer care, would be superior to either hospice care alone or a control intervention providing emotional support plus hospice care. We hypothesized that a problem solving intervention would increase caregivers' QOL and sense of mastery and would reduce caregiver burden. We also explored the effects of the interventions on caregiver coping.
MATERIALS AND METHODS
Setting and Sample
The sample for this study was drawn from consecutive admissions to a large nonprofit community-based hospice in the southeastern United States. A total of 480 patient/caregiver dyads were sought for this study. Inclusion Criteria: caregivers had to be providing care for adult patients with cancer, and both had to consent to participate, have at least a sixth grade education, be able to read and understand English, and achieve a minimum score of seven on the Short Portable Mental Status Questionnaire.21Exclusion criteria: caregivers were excluded if they were in active treatment for cancer themselves. Primary caregivers were identified by the hospice, and dyads were excluded if it was unclear who the primary caregiver would be or if patient performance status suggested that patients would not survive more than a few days.
Primary outcome measures
We assessed several indicators of caregivers' QOL, burden, and mastery as our primary outcomes ensuring that we measured domains of interest but without undue burden on already stressed caregivers. We measured multiple dimensions of burden and mastery because caregivers face multiple problems, including some that are specific to caregiving tasks, others to patient symptoms, and some that are relevant to the broad impact of caregiving on their lives. Specific indicators included:
Caregiver quality of life.
Caregiver QOL was assessed with the Caregiver Quality of Life Index-Cancer (CQOL-C), which is 35 items scored by using a five-point Likert-type scale that yields a single QOL score. Items assess the impact of cancer on the caregiver's mood, worry, sleep, daily life, family life, and other dimensions. The instrument is psychometrically sound and has been used with hospice caregiving samples.22
Burden of cancer symptoms.
Burden associated with patient cancer symptoms was assessed with the Memorial Symptom Assessment Scale (MSAS).23 The instrument includes 24 patient symptoms such as pain, lack of energy, diarrhea, and shortness of breath. We adapted the MSAS by asking caregivers to rate how distressing patient symptoms were to them (caregivers). Frequency and intensity were not assessed. This approach to measuring symptom-related burden has previously been used successfully in studies of dementia caregiving burden24 and allowed us to assess whether intervention altered caregivers' ability to adapt to problems that cannot always be eliminated.
General caregiver mastery.
General caregiver mastery was assessed by a six-item scale including caregivers' reports of their feelings of control and confidence in caregiving.25 The measure has good reliability, content validity, and concurrent validity.13
Burden and mastery specific to caregiving tasks.
Caregivers completed the Caregiver Demands Scale (CDS),26 which has 46 items that assess burden and mastery specific to caregiving tasks including assistance with meals, intimate care, treatments, and supervision of the patient. For each item, caregivers rated both how stressful the task was (burden) and their confidence in their ability to manage their stress related to this task (mastery).
Secondary outcome measures
We explored the impact of these interventions on coping responses, using the Brief COPE, a psychoeducational intervention that has been widely used in stress research.27 The Brief COPE Scale has 28 items scored by a 5-point Likert-type format. We scored the instrument for use of problem-focused coping (e.g., taking action) versus emotion-focused coping (e.g., blaming self). Problem-focused coping is generally associated with better well being in caregivers, whereas high levels of emotion-focused coping is associated with greater caregiver distress.24 Factor analysis has confirmed a two-factor structure including problem- and emotion-focused coping.28 Cronbach alpha for problem- and emotion-focused coping were acceptably high in the current sample at 0.74 and 0.62, respectively.
Short Portable Mental Status Questionnaire (SPMSQ).
The SPMSQ is a simple 10-item test of orientation, remote memory, and ability to complete a serial subtraction task.21 It is administered by the interviewer, scored 0–10 on the number of errors an individual makes, and measures levels of mental impairment.
Standard demographic data were collected on patients and caregivers including age, gender, education level, marital status, religion, occupation, cancer diagnosis of patient, and length of time since diagnosis.
This project was approved by the hospice's bioethics committee and the university's IRB. After approval, nurses and home health aides with hospice experience were hired onto the project and trained by the investigators for either intervention or data collection roles.
Newly admitted patient/caregiver dyads were identified by the nurse data collector each day. The data collector contacted the caregiver to arrange a visit by both the nurse and home health aide (HHA), during which consent of both patient and caregiver were obtained, the mental status of the dyad was assessed, and the functional status of the patient was evaluated. Baseline data were collected from subjects who met eligibility criteria. Although the sample could not be randomly selected, the dyads were randomly assigned to the three treatment conditions at baseline by using a computerized randomization procedure by telephone.
Caregivers in the control group received hospice standard care and participated in data collection. Data were collected by a team of research assistants; data were collected from caregivers by the nurse data collector and from the patient by the HHA data collector.
Caregivers in Treatment Group II received standard hospice care plus supportive visits from the intervention nurse and the HHA made on the same schedule as the COPE intervention group. The same amount of time was spent at each visit as with each COPE intervention visit (Group III). However, instead of providing caregivers with training on intervention techniques, the nurse spent this time providing individual support to the caregivers, discussing their feelings, fears, and relationships with their patients. The supportive visit intervention controlled for the effects of researcher attention, enabling the researchers to differentiate the effect of the intervention from the effect of the extra time and attention given to the caregivers. Interventionists were trained not to give advice to caregivers about managing problems and did not teach any structured problem-solving skills. Interventionists often responded to caregivers' questions with, for example, “What did your hospice nurse tell you to do in this situation?” During supportive visits, the HHA provided respite for the caregiver by remaining with the patient.
Caregivers in the COPE experimental group received standard hospice care and, in addition, were taught a problem solving method by the intervention nurse to assist them with assessing and managing patient symptoms. This coping intervention derives from the conceptual and research literature on problem solving training and therapy. The Family COPE model adapts these concepts to address the specific needs of families caring for persons with cancer at home.19 The model has four components: Creativity (viewing problems from different perspectives to develop new strategies for solving caregiving problems, e.g., “What could I do to distract Dad from his pain?”); Optimism (having a positive, but realistic, attitude toward the problem-solving process, e.g., “I believe I can help Dad with his pain.”). This includes communicating realistic optimism to the patient by showing both understanding and hope and involves patients, as much as possible, in planning; Planning (setting reasonable caregiving goals and thinking out, in advance, the steps necessary to reach those goals, e.g., “How can I get Dad ready for our family holiday party?”); and Expert information (what nonprofessionals need to know about the nature of the problem, when to get professional help, and what family caregivers can do on their own to deal with the problem, e.g., “Where can I go to get help with Dad's pain?”). Symptom assessment is included in the Expert Information element of the COPE model because accurate information about symptom characteristics and intensity are essential for knowing when to call for professional help as well as for planning what caregivers can do. The COPE intervention has been written into manual form and is available from the authors on request. The HHA, trained by hospice, provided respite so caregivers could focus on the intervention.
Training and treatment fidelity
A four day training session was provided to intervention nurses (one full-time and one back-up) to prepare them to offer COPE and supportive visit interventions. This training included how to implement the COPE method for managing patient problems, and how to use the Homecare Guide for Advanced Cancer,29 and allowed practice through role-playing. Both treatments were put into manual form and audiotapes of sessions were reviewed monthly to ensure that nurses followed the intervention protocols.
To avoid bias, a separate team of data collectors was hired and trained for the study. Training included screening dyads, obtaining consents, and administering the project instruments without offering information or assistance; these research assistants were blind to treatment conditions and had contact with intervention staff only at regularly scheduled staff meetings where individual cases were not discussed.
Caregiver data collection occurred within 72 hours of admission to hospice, 1 week postintervention (Day 16 of hospice admission) and 2 weeks postintervention (Day 30). The HHA stayed with the patient to assist as needed and to reassure the caregiver that the patient was not alone while the nurse remained with caregivers as they completed self-report instruments. If the patient declined during the study and was unable to self-report, then data were collected only from the caregiver. However, if the patient died, the caregiver was excused from the study because of grief.
We used ANOVA and chi-square tests to examine differences as a function of treatment group to ensure that the randomization procedure produced comparable groups. ANOVA and chi-square tests also were used to determine whether attrition had an impact on sample composition. Specifically, analyses were computed on baseline demographic characteristics, as well as baseline outcome measures, to determine whether participants who completed 30 days of data collection were initially different, as compared with persons who did not complete the entire longitudinal follow-up.
To examine longitudinal changes in outcomes for the caregivers across three times of measurement (baseline, Day 16, Day 30), we applied random effects regression models to our data.30–31Although this method provides the same basic information as traditional repeated measures analysis, in terms of the presence of group differences, longitudinal changes over time, as well as a potential group by time interaction, there are several advantages to this method of data analysis, in terms of the estimation of longitudinal changes in functioning. Chief among advantages of this method of data analysis is the ability to include persons for whom complete data are not available. This is directly relevant to the current case, given the large and expected amount of attrition across the 30-day follow-up period. Thus, analyses reported include persons who contributed any data for the three measurement points.
For each dependent variable, four effects were estimated corresponding to an intercept or average value for all groups at the midpoint of the study, a group effect that examines whether differences at the intercept vary as a function of intervention group, longitudinal changes over time, and a Group X time interaction. The latter term is important here because it allows us to determine whether trajectories of change vary systematically across intervention groups. In cases where the Group X time interaction is statistically significant, separate random effects models estimating longitudinal changes over time were computed for each of the intervention groups to decompose the effects.
Baseline Demographic Characteristics
Among 354 caregiver–patient dyads who were eligible to participate in the study, 329 were successfully randomized to the three treatment groups: standard care (n = 109), standard care + support (n = 109), and standard care + COPE (n = 111). Twenty-five dyads were not randomized to a treatment condition because either the patient or the caregiver failed the initial screening. The demographic characteristics of participants who were randomized to the three intervention groups are shown in Table 1. The caregivers were overwhelmingly female, approximately 60 years of age, and the mean educational attainment was slightly greater than a high school graduate. None of the comparisons across intervention groups for caregiver demographic characteristics was statistically significant (P > 0.10). Among patients examined in the current study, the mean age was approximately 70 years, and the majority of them were male, with educational attainment slightly lower than caregivers but approximately equal to a high school graduate. As with caregiver analyses, no demographic characteristics showed significant differences across groups (P > 0.10). Lack of differences across the three groups indicates that the randomization was successful, from the standpoint of creating groups that were demographically comparable.
Table 1. Baseline Demographic Characteristics By Intervention Group
Standard care + support
Standard care + COPE
M: mean; SD: standard deviation.
Caregiver yrs of
Patient yrs of
Longitudinal Attrition Analyses
Before longitudinal analyses are described, it should be noted that there was substantial attrition across all three intervention groups across three times of measurement. Specifically, for the standard care Group, 40 (36.7%) persons were interviewed at Day 30. Similar rates of attrition were observed for the standard care + support group (n = 32; 29.4%), and the standard care + COPE participants (n = 31; 27.9%). However, rates of attrition did not vary systematically across the three intervention groups (chi square (2) = 2.26, P = 0.32). Most commonly, attrition was due to patient decline (29%) or death (21%), and the caregiver feeling overwhelmed (23%), which was largely associated with patients' worsening condition.
To examine whether those persons who completed the longitudinal follow-up were comparable to those persons who were initially randomized to the treatment groups, we examined the effect of longitudinal attrition on baseline demographic characteristics and performance on outcome measures. Among demographic characteristics, a significant effect of attrition for patient age was observed (F[1,322] = 8.09, P < 0.01), with those who completed the longitudinal assessment being initially older (mean age = 73.4 years), as compared with those persons who were lost to follow-up (mean age = 69.4 years). No other comparisons for age, years of education, or gender composition were statistically significant (P > 0.10). We then computed attrition X treatment group ANOVAs on outcome measures in the current analysis. None of the effects of attrition by group were statistically significant (P > 0.10).
Random Effects Models of Change
Table 2 displays results for random effects models for each of seven primary outcome measures. There were significant interactions, or trends in that direction, between time and intervention group for caregiver QOL (P = 0.054), burden related to patient symptoms (P = 0.001), and caregiving task burden (P = 0.021). For caregiving task burden, there was also a main effect of time (P = 0.014) with increases in burden scores being seen than more time.
Table 2. Summary of Random Effects Models Over the Follow-Up Group, as a Function of Intervention Group Status
Unstandardized estimates from the random effects models are shown
Caregiver quality of life
Group X Time
Group X Time
Caregiving task burden
Group X Time
Group X Time
Caregiver task mastery
Group X Time
Group X Time
Group X Time
To decompose significant interactions between time and intervention group seen above, separate random effects models for pairs of intervention groups were run for caregiver QOL, symptom burden, and caregiving task burden. Specifically, analyses compared the usual care to support group, as well as the usual care to COPE group. In this case, the analyses again examine whether a significant time by treatment group interaction is present that would indicate that the intervention group is changing at a different rate than the usual care control group. The results of this analysis indicated that the treatment group by time interactions were not statistically significant for the comparison between usual care and support condition for caregiver QOL (estimate = −0.06, standard error [SE]) = 0.09, P = 0.512), symptom burden (estimate = 0.13, SE = 0.08; P = 0.118) or caregiving task burden (estimate = 0.01, SE = 0.01, P = 0.542). By contrast, for the comparison between the usual care group and the COPE intervention group, the group by time interactions for caregiver QOL (estimate = 0.096, SE = 0.049, P = 0.042), symptom burden (estimate = −0.14, SE = 0.04, P < 0.001), and caregiving task burden (estimate = 0.01, SE = 0.01, P = 0.04) were all statistically significant. The results of this analysis are graphically portrayed in Figure 2. For caregiver QOL, only the COPE intervention group showed statistically significant improvements in QOL ratings over time (estimate = 0.16, SE = 0.07, P = 0.033), whereas the usual care group experienced no significant change over time (estimate = 0.02, SE = 0.06, P = 0.718). As seen in Figure 2, the ratings improved by almost 10% over baseline levels. For symptom burden, a similar pattern was present with significant improvements being seen in the COPE intervention group (estimate = −0.28; estimate = 0.07; P < 0.001) but not for the usual care group (estimate = 0.01; SE = 0.01; P = 0.847). As shown in Figure 2, the COPE group improved by 30% while the usual care group showed a slight increase in scores. Finally, none of the time effects for the usual care (estimate = 0.01; SE = 0.01; P = 0.375) or COPE intervention group (estimate = 0.01; SE = 0.01; P = 0.150) were statistically significant for the caregiving task burden measure; however, the source of interaction was likely due to the finding that the COPE group improved over time, whereas the usual care group exhibited increased burden scores.1
The COPE intervention was uniquely effective in improving caregivers' overall QOL and in decreasing burden related to patients symptoms and caregiving tasks, which are essential goals of hospice and palliative care.14–15 It would have been helpful to be able to look at which aspect of QOL was affected by the intervention, but this was not possible using a scale that provides only a total QOL score; this should be addressed in future studies. The burden (distress) from patient symptoms and caregiver tasks also was significantly improved by the intervention. In a practical sense, this means that caregivers were better able to tolerate patient symptoms such as pain and constipation, or tasks such as providing personal care, even when hospice care did not eliminate symptoms or remove the need for care. This study showed that teaching the COPE method is a way of improving caregiver well being and enhancing existing hospice care. It should be noted that many patients with advanced cancer do not receive hospice care, and it is possible that this intervention also may be helpful to them and their caregivers; further study is warranted.
The caregivers' feelings of mastery were not affected by the intervention. Mastery may be difficult to improve in hospice given the very demanding nature of the caregiving involved. The COPE intervention is a well developed theory-based intervention that can be immediately translated to the bedside. Evidence of its usefulness, therefore, is very relevant. Structured instruction in coping skills is one way to equip caregivers with skills to deal with a situation in which they may be overwhelmed with demands and are faced with multiple ongoing stressors, and the impending death of a loved one. It was also of interest that this three-session coping skills intervention was superior to a control intervention that provided the same number of sessions of emotional support, and that this emotional support condition showed no significant effects on any of the dependent measures compared with hospice care alone. Emotional support alone may be appreciated by caregivers, but it appears to be ineffective in improving caregiver QOL or caregiver burden due to patient symptoms.
Although the COPE intervention was found to be effective for caregivers of hospice patients with cancer, it should be noted that a growing proportion of hospice patients have other life-limiting conditions including end-stage dementia and cardiac and pulmonary disease. When this study was in the planning stage, approximately 80% of patients in this hospice had a cancer diagnosis; currently national figures show that only about 50% of hospice patients have cancer.20 This suggests a need to replicate this study with other hospice populations. We recommend future research that replicates our study of hospice patients with cancer. In addition, studies should be designed to evaluate effectiveness of the COPE intervention with caregivers of hospice patients with noncancer diagnoses.
Because this intervention has been written into a manual, it is replicable and can be readily used by professionals providing end-of-life care in a variety of contexts. We hope that our work will contribute to the development of greater evidence-based practice for family caregivers at the end of life. There has not been a strong evidence base for psychosocial care at the end of life to date, and our project demonstrates that it is feasible to conduct a randomized trial even in the difficult context of hospice care.
Accrual was difficult, and attrition was high in this study as might be expected with this study population. Nationally, 34.7% of hospice patients die within 7 days or less of admission to hospice.20 Detailed discussion of the accrual and attrition problems encountered in this project are published elsewhere.32 The similarity of the groups at baseline supports the success of the randomization procedure, and there was no differential attrition across groups. Only one variable, younger patient age, predicted attrition. It is unclear why younger persons had slightly higher attrition, but this finding suggests that advanced age does not preclude participation in a trial of this type.
Future studies can be modified to reduce measurement burden on caregivers. For example, studies could be developed that include only those variables that demonstrated an effect from the intervention. In addition, decreasing the duration of the study from 30 to 20 days would decrease attrition.
The current results are especially impressive given that, with a maximum of three training sessions, we were able to improve caregiver QOL and decrease burden among families who were receiving hospice services. Hospice provides a great deal of support for families, and caregivers generally report very high satisfaction with hospice care.17 We think it is likely that the impact of the coping skills intervention would provide even greater impact when compared with family caregivers who are not receiving the support of hospice care.
Although research is difficult in the setting of hospice, this study demonstrated the viability of conducting clinical trials with patients and caregivers near the end of life and provided evidence of effectiveness of this intervention in improving important indicators of caregiver well being. Although this structured caregiver skill-training intervention is in a manual and is readily translatable to the bedside, further research is needed with both cancer and noncancer populations to assess issues such as the longer-term impact of such intervention on the family caregiver, e.g., during bereavement.
Current results are consistent with an impressive body of evidence showing that family caregivers need much more than just our concern and emotional support, and that caregivers can show substantial and sustained benefits from intervention when given structured psychosocial intervention.5, 33 It is important that we do more to develop and implement evidence-based interventions for caregivers. Families who sacrifice so much for the care of their loved ones deserve access to effective interventions, and research can help establish effective ways to assist caregivers that go beyond clinical lore in the delivery of psychosocial care.
The authors acknowledge the many contributions of LifePath Hospice and Palliative Care, the Center for Hospice, Palliative Care, and End-of-Life Studies at the University of South Florida, and project staff, including Melissa Leggatt, A.A., Leah Sisler, MS, ARNP, Ann Holzheimer, MS, ARNP, Phyllis Ruvo, RN, Kim Davis, RN, Barbara Diaz, HHA and Fran Young, HHA.